Natural resources and capital structure

Transcription

1 BOFIT Discussion Papers Sanna Kurronen Natural resources and capital structure Bank of Finland, BOFIT Institute for Economies in Transition

2 BOFIT Discussion Papers Editor-in-Chief Zuzana Fungáčová BOFIT Discussion Papers 10/ Sanna Kurronen: Natural resources and capital structure ISBN , online ISSN , online This paper can be downloaded without charge from Suomen Pankki Helsinki 2016

3 BOFIT- Institute for Economies in Transition Bank of Finland BOFIT Discussion Papers 10/ 2016 Contents Abstract Introduction Capital structure of resource firms Data and methodology Data description Methodological strategy Results Variance decomposition Regression results Conclusions References Appendices Appendix 1 Equity indices included Appendix 2 Data description and sources Appendix 3 Variable means by country

4 Sanna Kurronen Natural resources and capital structure Sanna Kurronen Natural resources and capital structure Abstract This paper examines the effect of natural resources on capital structure of the firm. Using an extensive dataset of listed firms in 70 countries, we show that firms operating in resource extraction industries have less debt and that that debt tends to have a longer maturity than that of other non-financial firms. Moreover, non-resource firms in resource-dependent countries are found to be less indebted than their counterparts in other countries. The results suggest that the very fact of a firm s location in a resource-dependent country may be an overlooked country-specific determinant of firm capital structure and that financial institutions in resource-dependent countries may play a role in exacerbating a nation s resource curse. JEL classifications: G32, O13, Q32. Keywords: resource dependence, capital structure, panel data. Sanna Kurronen, orcid.org/ University of Helsinki, Economicum, Arkadiankatu 7. sanna.kurronen@gmail.com. Acknowledgements I would like to thank Bank of Finland s Institute for Economies in Transition (BOFIT) for providing the opportunity to use their data sources and also present my work in a BOFIT seminar. The many valuable comments I received there and later on have helped to improve this work considerably. 4

5 BOFIT- Institute for Economies in Transition Bank of Finland BOFIT Discussion Papers 10/ Introduction In countries highly dependent on their mineral resource sectors, the failure to diversify industrial activity is sometimes characterized as a resource curse. It is not clear, however, whether a resource curse is merely the natural outcome of organizing an economy around its resource sector based on factor endowments. In this paper, we consider the link between resource sector and finance. Given the dominance of the resource sector in cursed countries, we presume that financial institutions there are focused on meeting the needs of the resource sector. Kurronen (2015) notes that resourcedependent economies tend to extend less domestic credit to the private sector and rely more heavily on market-based financial instruments than their non-resource-dependent counterparts. Here, we extend the discussion to firm level and consider how capital structure of a firm differs from other firms when it operates directly in the resource sector or otherwise happens to be located in a resource-dependent country. Our hypothesis is that financial sectors in resource-dependent countries are geared to serving large, well-known resource firms with considerable tangible assets. These conditions result in a financial infrastructure that may be challenging for small firms and emerging industries. We test our hypothesis using an extensive micro-level dataset containing financial data for listed firms in 70 countries. Listed firms in general are larger on average than non-listed firms and enjoy easier access to external finance (Baum et al., 2011). We contribute to the existing literature in two ways. First, we consider how the capital structure of a resource firm might differ from firms in other sectors. We present empirical evidence covering a wide range of countries that suggest resource firms tend to have less debt than other non-financial firms and that that debt has a longer maturity. Second, we show that other firms in resource-dependent countries are less indebted than their counterparts in other countries. For this reason, we argue that mere location in a resource-dependent country is a country-specific determinant of firm capital structure. The remainder of this paper consists of four sections. Section 2 introduces the related literature. Section 3 discusses the data and methodology. Section 4 presents the empirical results. Section 5 concludes. 5

6 Sanna Kurronen Natural resources and capital structure 2 Capital structure of resource firms Contrary to the classic assumption of Modigliani and Miller (1958), firms do not always choose debt levels optimal to their needs. The literature shows, for example, that, due to supply frictions, observed capital structures differ from those demanded by the firms (Faulkender and Petersen, 2006). Beck (2011) makes a similar assertion based on survey data of firms in resource-dependent countries. Recent literature highlights firm- and industry-specific factors affecting the capital structure of firms. Frank and Goyal (2009) show that leverage tends to increase with firm size and more tangible assets. Lower leverage, in turn, is related to higher profitability and high market-to-book ratios. They also find evidence that firms increase leverage when anticipate high inflation. These results are not unambiguous, however. Considering data for nine Eastern European countries, Jõeveer (2013) finds that firms with a high share of tangible assets have lower leverage. Fan et al. (2012) demonstrate that country-specific factors are more important in determining firm capital structure than the particular industry in which the firm does business. They also find that legal systems originating in common law are associated with lower debt ratios, whereas higher development level, higher corruption and the existence of an explicit bankruptcy code are related to higher debt ratios. Higher debt ratios are also observed in countries where the tax benefit of leverage is positive. This study further notes that debt maturity tends to be longer in countries with common law legal origins and shorter in more corrupt countries and in countries with large government bond markets. Specifically, the authors suggest that suppliers of capital influence the debt-ratio choices of firms. They find that leverage is higher in countries with deposit insurance, suggesting that the role of banking industry is important. Jõeveer (2013) finds evidence for emerging countries that a large presence of foreign banks and high level of bank concentration coincide with lower leverage of firms. Holmstrom and Tirole (1997) argue that lending to large firms is less vulnerable to credit supply shocks than lending to smaller or riskier firms. Further, borrowers facing relatively high agency costs are the first to face limitations in access to finance in a flight to quality" (e.g. Bernanke et al., 1996). Given that resource firms are typically large, well-known and possess considerable tangible assets, we would expect a certain degree of immunity to supply shocks and easier access to finance for resource firms than other firms in resource-dependent countries. 6

7 BOFIT- Institute for Economies in Transition Bank of Finland BOFIT Discussion Papers 10/ 2016 Recent discussions in structural economics highlight the evolving role of the financial sector at various stages of economic development. As economies develop, they tend to become increasingly reliant on market-based finance. Moreover, a country s deviation from its optimal financial structure is reflected in depressed levels of economic activity (Demirguc-Kunt et al., 2011). Lin et al. (2009) observe that the optimal mix of banks and markets or big and small banks depends on the economy s factor endowments. The relative composition of labor, capital and natural resources define the optimal structure for production, while the production structure defines the optimal financial sector. Capital- intensive countries tend to have big production firms and are thus better served by a market-based financial system or big banks. Labor-intensive economies, in contrast, have smaller firms better served by small, local banks. Unfortunately for our purposes, the authors merely acknowledge natural resources as an initial factor endowment without delving deeper into the specific role of natural resources. Engerman and Sokoloff (2002) and Acemoglu et al. (2001) discuss colonial endowments. They note that colonies built around extractive industries or agriculture with large returns to scale tended to have weak property rights. In colonies settled by large groups of immigrants, in contrast, property rights tended to be stronger and levels of education and financial and economic development higher. As a result, beneficial institutions could not be said to be exogenously determined. To the best of our knowledge, no paper in the literature investigates the capital structure of resource firms or the capital structure of firms in resource-dependent countries using microlevel data. In contributing to the existing literature, our hypothesis is that large resource assets lead to a resource-dependent economy with a financial sector geared to serving large resource firms. Smaller firms and emerging industries thus lack adequate access to financial services, thereby exacerbating the resource curse. 3 Data and methodology 3.1 Data description Using firm data from Bloomberg, we gather financial data from companies included in the main equity indices of 73 countries over the period For the largest equity market, the US, we use firms in the S&P500. A list of all the equity indices used appears in Appendix 1. 7

8 Sanna Kurronen Natural resources and capital structure Our approach omits fully state-owned companies, which obviously play huge roles in many resource-rich countries. The problem is that financial information on such companies is often quite limited, which makes them anyway difficult to include in the data (Wolf, 2009). We also limit the data to non-financial firms and countries with observations for at least three firms. We remove observations with missing values on debt or assets and trim the data by excluding observations where book leverage exceeds four times the median absolute deviation from the median. 1 Our final sample consists of 4,319 non-financial firms over seven years and 25,373 firm-year observations from 70 different countries of domicile. We measure capital structure with commonly used indicators (Fan et al., 2012). Book leverage, or more precisely, short-term and long-term interest-bearing debt to total assets is used as the main indicator of company leverage as this is the most available indicator on leverage. While ratios based on market values might be more relevant, managers focus on book leverage because debt is better supported by assets in place than by growth opportunities (Myers, 1977). Book leverage is also preferred because financial markets fluctuate considerably (as evidenced during our sample period). We use market leverage, i.e. short- and long-term interest-bearing debt to total market value of the firm as an alternative measure of leverage. To provide a more thorough picture of the capital structure of firms, we separately consider the ratio of short- and long-term debt to assets and the share of long-term debt to total debt as a measure of debt maturity. As the investments of resource firms tend to be bulky, we expect them to have debt with longer maturity than non-resource firms (Berglof and Lehmann, 2009). As our firm-specific control variables, we use common measures of firm size, tangibility and profitability (see e.g. Titman and Wessels, 1988). Firm size is measured by taking a natural logarithm of the US dollar value of total assets. As a measure of tangibility, we use the amount of property, plant and equipment relative to total assets. Profitability is measured by cash from operations to total assets as it describes the capability of the firm to generate cash to finance investments. We also use market-to-book ratio as an additional firm-specific variable to describe growth opportunities. Our country-specific control variables are mostly taken from the World Bank World Development Indicators (WDI). We use variables that the literature finds significantly related to capital structure measures, i.e. GDP growth rate, inflation, bank concentration, domestic lending 1 As we are very careful in removing outliers as the tails of distribution could contain valuable information, our approach initially excludes only 56 or 0.2% of firm-year observations. Thereafter, we test the robustness of the results with more restricted samples. 8

9 BOFIT- Institute for Economies in Transition Bank of Finland BOFIT Discussion Papers 10/ 2016 to private sector, stock market turnover, corruption and profit tax rate. 2 We also include three binary variables: developed to indicate a country was classified as high income country by World Bank in 2008, deposit insurance to show the country has some sort of deposit insurance scheme, and common law to highlight common law origins of the legal system. Credit rating is taken from Standard and Poor s ratings as of The summary statistics are presented in Table 1. 3 Both firm and country variables and their sources are described in detail in Appendix 2. The market variables in Table 1, the trading volume of equity markets and market-to-book ratio suffer extensively from missing values. We omit them from our regressions whenever the estimated coefficient for the variable in question is insignificant to reduce the loss of observations. We do the same with bank concentration, credit to private sector and tangibility. Table 1 Summary statistics of selected variables Statistic n Mean St. Dev. Min. Median Max. Book leverage 25,373 0,24 0,18 0,00 0,23 1,02 Market leverage 23,506 0,25 0,22 0,00 0,20 1,07 Maturity 25,373 0,53 0,36 0,00 0,62 1,00 St debt 25,373 0,09 0,12 0,00 0,05 1,00 Lt debt 25,373 0,15 0,15 0,00 0,12 1,02 Size 25,230 6,75 2,85-9,39 7,13 13,59 Tangibility 23,018 0,34 0,24 0,00 0,31 1,02 Profitability 25,223 0,09 0,12-3,32 0,08 1,68 Market-to-book 23,509 1,44 1,34 0,02 1,04 28,32 Corruption 25,373-0,43 1,05-2,53-0,08 1,28 CPI 25,251 4,29 4,07-4,86 3,27 40,64 Concentration 24,280 0,61 0,26 0,07 0,60 1,00 Private credit 24,135 1,09 0,58 0,11 1,13 2,24 Market activity 23,647 82,32 90,77 0,02 58,09 952,67 GDP growth 25,373 3,34 3,87-14,81 2,96 19,59 GDP/cap 25,369 24, , ,17 15, , Profit tax 25,366 0,39 0,14 0,11 0,37 1,19 Variables: Book leverage Total long- and short-term interest bearing debt to total assets; Market leverage Total long- and short-term interest bearing debt to market value of the firm; Maturity Long-term debt total debt; St debt Short-term interest bearing debt to total assets; Lt debt Long-term interest bearing debt to total assets; Size Natural logarithm of assets in US dollars, millions; Tangibility Fixed assets to total assets; Profitability Cash from operations to total assets; Market-tobook Market value to total assets; Corruption Corruption, high value indicates more corrupt, CPI Consumer price inflation, %, year-on-year; Concentration The share of assets of the three largest banks of total bank assets; Private credit Domestic credit to private sector, % of GDP; Market activity Stock market turnover, % of GDP; GDP growth Annual real GDP growth rate, %; GDP/cap Gross domestic product in US dollars per capita; Profit tax Profit tax, % of commercial profits. 2 For some countries, we have only one observation for profit tax rate in As tax rates generally do not fluctuate much, we use this observation for all years. In any case, when we test the results without the indicator they remain very similar. For corruption, we have inverted the scale of original data for higher values to indicate more corrupt. 3 Variable means by country are listed in Appendix 3. 9

10 Sanna Kurronen Natural resources and capital structure We classify resource firms as firms that have GICS classifications in the industrial categories Metals & Mining and Oil & Gas Exploration & Production or its sub-industry categories Oil & Gas Drilling, Integrated Oil & Gas or Coal & Consumable Fuels. This gives us 580 individual firms and 3,501 firm-year observations. Resource-dependent countries are defined as countries where minerals account for more than 40% of total exports on average during the sample period (Nili and Rastad, 2007). Because our purpose is to establish whether or not a given country s competitiveness is based largely on minerals, we use mineral exports to total exports as our indicator of resource dependence. The alternative measure of mineral exports in excess of 10% of GDP is overbroad here as it captures countries such as Estonia, which has a very large export sector but modest resource endowments. Including such countries as resource-dependent would distort our findings. Countries where minerals share of total exports exceeds 40% in our sample include Australia, Bahrain, Chile, Colombia, Egypt, Kazakhstan, Kuwait, Nigeria, Norway, Oman, Peru, Qatar, Russia, Saudi Arabia, South Africa, United Arab Emirates and Venezuela. However, as WDI data omits diamond producers, we follow Kurronen (2015) and add diamond exports data to major diamond producers where data was available. Thus, Botswana was included in the group of resource-dependent countries so we have 18 countries out of 70. The correlation matrix in Table 2 shows that more profitable firms have less debt and that bigger and more tangible firms use more debt, which is in line with Frank and Goyal (2009). Longer debt maturity is associated with larger firm size, jurisdictions with common law legal origins, lower rates of corruption and greater economic development. High rates of GDP growth, inflation and corruption seem to coincide with shorter debt maturity. Among our control variables, corruption seems to be highly correlated with other explanatory variables. In particular, it is highly and negatively correlated with level of economic development, credit rating and level of bank credit to private sector. 10

11 BOFIT- Institute for Economies in Transition Bank of Finland BOFIT Discussion Papers 10/ 2016 Table 2 Correlation matrix Book leverage 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,06 0,00 0,00 0,00 0,08 0,00 0,00 0,22 0,00 0,00 0,01 2 Market leverage 0,79 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,80 0,00 0,00 0,00 0,63 0,00 0,00 0,00 3 Maturity 0,32 0,19 0,00 0,00 0,00 0,00 0,00 0,00 0,01 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 4 St debt 0,58 0,54 0,39 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,28 0,00 0,00 0,00 0,00 0,00 0,00 0,85 0,00 5 Lt debt 0,76 0,55 0,69 0,08 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 6 Size 0,13 0,08 0,43 0,18 0,30 0,21 0,00 0,00 0,36 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,58 0,00 0,00 7 Tangibility 0,19 0,16 0,21 0,02 0,25 0,01 0,00 0,00 0,00 0,00 0,00 0,05 0,00 0,00 0,00 0,00 0,00 0,00 0,08 0,00 0,00 8 Profitability 0,21 0,31 0,06 0,25 0,07 0,08 0,10 0,00 0,99 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,96 0,09 0,00 9 Market-to-book 0,17 0,42 0,10 0,12 0,12 0,04 0,08 0,26 0,00 0,00 0,01 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,53 0,00 10 Resource firm 0,06 0,08 0,02 0,03 0,05 0,01 0,17 0,00 0,04 0,00 0,00 0,00 0,00 0,02 0,03 0,00 0,00 0,91 0,00 0,00 0,00 11 Resource country 0,08 0,10 0,04 0,05 0,06 0,10 0,07 0,04 0,05 0,03 0,29 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 12 Corruption 0,01 0,05 0,41 0,29 0,25 0,51 0,09 0,06 0,02 0,06 0,01 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 13 CPI 0,03 0,04 0,23 0,15 0,15 0,38 0,01 0,03 0,05 0,03 0,15 0,54 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,15 0,00 14 Concentration 0,03 0,07 0,05 0,01 0,04 0,04 0,05 0,03 0,15 0,05 0,18 0,20 0,10 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 15 Private credit 0,05 0,00 0,27 0,14 0,17 0,50 0,17 0,03 0,03 0,02 0,32 0,66 0,54 0,04 0 0,00 0,00 0,00 0,00 0,00 0,00 16 Market activity 0,01 0,12 0,25 0,18 0,13 0,47 0,11 0,08 0,14 0,01 0,18 0,46 0,32 0,21 0,63 0,00 0,00 0,00 0,00 0,52 0,00 17 Common law 0,04 0,14 0,33 0,28 0,17 0,31 0,03 0,10 0,11 0,12 0,03 0,57 0,16 0,08 0,44 0,51 0,00 0,00 0,00 0,00 0,00 18 GDP growth 0,03 0,12 0,23 0,15 0,16 0,25 0,03 0,03 0,17 0,02 0,08 0,45 0,31 0,30 0,36 0,14 0,24 0,00 0,00 0,00 0,00 19 Developed 0,01 0,00 0,34 0,26 0,21 0,48 0,07 0,04 0,03 0,00 0,08 0,83 0,49 0,15 0,54 0,45 0,45 0,50 0,09 0,00 0,00 20 Deposit insurance 0,04 0,12 0,11 0,04 0,08 0,00 0,01 0,00 0,17 0,03 0,44 0,05 0,06 0,06 0,15 0,04 0,09 0,30 0,01 0,00 0,00 21 Profit tax 0,06 0,07 0,07 0,00 0,07 0,12 0,06 0,01 0,00 0,04 0,22 0,03 0,01 0,12 0,06 0,00 0,02 0,05 0,12 0,13 0,00 22 Rating 0,02 0,14 0,32 0,25 0,18 0,52 0,06 0,08 0,13 0,09 0,08 0,79 0,57 0,07 0,61 0,56 0,46 0,18 0,67 0,19 0,08 Notes: Pearson correlation coefficient in lower triangle and corresponding p-values in the upper triangle.variables Book leverage Total long- and short-term interest bearing debt to total assets; Market leverage Total long- and short-term interest bearing debt to market value of the firm; Maturity Long-term debt total debt; St debt Short-term interest bearing debt to total assets; Lt debt Long-term interest bearing debt to total assets; Size Natural logarithm of assets in US dollars, millions; Tangibility Fixed assets to total assets; Profitability Cash from operations to total assets; Market-to-book Market value to total assets; "Resource firm" Binary variable for 1=resource firm; "Resource country" Binary variable with 1=Resource-dependent country; Corruption Corruption, high value indicates more corrupt, CPI Consumer price inflation, %, year-on-year; Concentration The share of assets of the three largest banks of total bank assets; Private credit Domestic credit to private sector, % of GDP; Market activity Stock market turnover, % of GDP; "Common law" Binary variable with 1=common law legal origins; GDP growth Annual real GDP growth rate, %; Developed Binary variable with 1=developed country; "Deposit insurance" Binary variable for deposit insurance with 1=deposit insurance scheme; Profit tax Profit tax, % of commercial profits; "Rating" S&P country credit rating. 11

12 Sanna Kurronen Natural resources and capital structure Based on our data, resource firms have lower debt levels and that carry debt of longer maturity than other firms. The difference is statistically significant for both leverage variables at 1% level based on the Welch Two Sample t-test (Table 3). Due to the volatile period around the global financial crisis, we also check the variables for each year separately to discover any anomalies that might drive our results. The result for significant difference in both book and market leverage is valid for each year in our sample except for 2013, where we find no significant difference for resource firms and other firms. The result on debt maturity is not as strong; we find statistically significant difference for individual years between the two groups only for 2012 and 2013 at the 10% and 5% significance levels, respectively. When dividing our sample by country groups, we find the result of significant difference in leverage between resource and non-resource firms robust for rich countries and resource-dependent countries. However, we find no significant difference in developing countries for book leverage for resource and non-resource firms. The leverage for resource firms is clearly higher in developing countries than in developed countries. For developed countries, we find no significant difference in debt maturity for resource and non-resource firms. Summary statistics are presented in Table 3 for various country groups. Table 3 also shows that resource firms have more tangible assets than other firms in our data except such firms in resource-dependent countries. Most empirical evidence has shown (Frank and Goyal 2009) that, like the resource firms in our data, firms with more tangible assets are expected to have more, not less, debt. This finding might be due to the volatile end product prices of raw materials, which heighten uncertainty of cash flow for resource firms, despite their observed asset tangibility. Resource firms are also no larger in terms of assets than other firms except in resource-dependent countries. This finding could be explained by the fact that our sample consists only of firms included in the main equity index of each country. We find no difference in profitability for resource firms and other firms. We confirm the findings with US data, where the differences in country-specific factors do not disturb the analysis. As US financial markets have size and depth to service the needs of the firms, we expect firm capital structure in the US to well reflect the demand for capital. Within our sample of 420 non-financial US firms, 41 are classified as resource firms. The results in Table 3 are robust with the cross-country data. 12

13 BOFIT- Institute for Economies in Transition Bank of Finland BOFIT Discussion Papers 10/ 2016 Table 3 Summary statistics of the firm variables by groups Resource firms Non-resource firms N Mean St. Dev. Min. Median Max. N Mean St. Dev. Min. Median Max. Welch t-test p-value All countries Book leverage 3,501 0,21 0,17 0 0,2 0,94 21,872 0,24 0,18 0 0,23 1,02 <0.01 Market leverage 3,193 0,21 0,2 0 0,15 0,95 20,313 0,26 0,22 0 0,21 1,07 <0.01 Maturity 3,501 0,55 0,38 0 0, ,872 0,53 0,36 0 0,61 1 0,01 Size 3,494 6,71 3,38-9,39 7,25 13,08 21,736 6,75 2,75-8,9 7,1 13,59 <0.01 Tangibility 3,001 0,44 0,26 0 0,44 1,02 20,017 0,33 0,23 0 0,29 1 <0.01 Profitability 3,484 0,09 0,15-3,32 0,09 1,59 21,739 0,09 0,11-2,45 0,08 1,68 0,99 Developed countries Book leverage 1,825 0,19 0,15 0 0,18 0,91 11,378 0,25 0,17 0 0,24 1,01 <0.01 Market leverage 1,697 0,18 0,18 0 0,15 0,89 10,883 0,26 0,21 0 0,22 1,07 <0.01 Maturity 1,825 0,64 0,37 0 0, ,378 0,65 0,33 0 0,77 1 0,42 Size 1,818 7,75 2,36-2,33 7,94 13,08 11,242 8,11 2,06 0,39 8,4 13,59 0,01 Tangibility 1,716 0,49 0,28 0 0,51 1,02 10,868 0,3 0,22 0 0,26 1 <0.01 Profitability 1,824 0,09 0,15-3,32 0,1 1,59 11,357 0,09 0,09-2,45 0,09 1,1 0,2 Developing countries Book leverage 1,676 0,25 0,19 0 0,23 0,94 10,494 0,24 0,19 0 0,22 1,02 0,12 Market leverage 1,496 0,24 0,23 0 0,17 0,95 9,430 0,26 0,24 0 0,2 0,99 0,01 Maturity 1,676 0,44 0,35 0 0, ,494 0,4 0,35 0 0,37 1 <0.01 Size 1,676 5,57 3,93-9,39 6,47 12,95 10,494 5,3 2,66-8,9 5,21 11,76 <0.01 Tangibility 1,285 0,38 0,21 0 0,38 0,94 9,149 0,36 0,24 0 0,32 1 <0.01 Profitability 1,660 0,09 0,15-1,75 0,08 0,77 10,382 0,08 0,13-1,62 0,08 1,68 0,2 Resource-dependent countries Book leverage 716 0,18 0,16 0 0,15 0,91 3,697 0,21 0,18 0 0,2 0,93 <0.01 Market leverage 668 0,16 0,17 0 0,11 0,89 3,407 0,21 0,2 0 0,17 0,99 <0.01 Maturity 716 0,53 0,37 0 0,65 1 3,697 0,49 0,38 0 0,56 1 0,01 Size 716 6,8 2,32-0,98 6,71 12,95 3,697 6,03 2,08-0,71 6,1 11,41 <0.01 Tangibility 697 0,38 0,24 0 0,38 0,89 3,553 0,38 0,24 0 0,35 0,98 0,63 Profitability 703 0,1 0,15-0,91 0,1 1,59 3,615 0,1 0,12-1 0,09 1,1 0,87 Non-resource countries Book leverage 2,785 0,22 0,18 0 0,21 0,94 18,175 0,25 0,18 0 0,24 1,02 <0.01 Market leverage 2,525 0,23 0,21 0 0,17 0,95 16,906 0,27 0,23 0 0,22 1,07 <0.01 Maturity 2,785 0,55 0,38 0 0, ,175 0,54 0,36 0 0,62 1 0,04 Size 2,778 6,68 3,61-9,39 7,42 13,08 18,039 6,9 2,85-8,9 7,41 13,59 <0.01 Tangibility 2,304 0,46 0,26 0 0,46 1,02 16,464 0,32 0,22 0 0,28 1 <0.01 Profitability 2,781 0,09 0,14-3,32 0,09 0,77 18,124 0,09 0,11-2,5 0,08 1,68 0,79 US Book leverage 279 0,2 0,1 0 0,21 0,47 2,584 0,24 0,16 0 0,24 1,01 <0.01 Market leverage 267 0,19 0,13 0 0,17 0,73 2,487 0,21 0,17 0 0,16 0,96 0,1 Maturity 279 0,89 0,23 0 0,98 1 2,584 0,8 0,29 0 0,91 1 <0.01 Size 279 9,29 1,68-0,76 9,26 12,76 2,584 9,29 1,25 1,54 9,24 13,59 0,04 Tangibility 271 0,67 0,19 0,01 0,72 0,96 2,376 0,23 0,19 0 0,16 0,9 <0.01 Profitability 279 0,13 0,08-0,14 0,13 0,41 2,583 0,12 0,09-2,45 0,11 0,52 0,02 Variables: Book leverage Total long- and short-term interest bearing debt to total assets; Market leverage Total long- and shor- term interest bearing debt to market value of the firm; Maturity Long-term debt total debt; Size Natural logarithm of assets in US dollars, millions; Tangibility Fixed assets to total assets; Profitability Cash from operations to total assets. Welch two-sample t-test will null hypothesis: no difference in means. 13

14 Sanna Kurronen Natural resources and capital structure Table 3 also highlights the fact that non-resource firms in resource-dependent countries seem to have less debt than their counterparts in other countries. This could be due to different industrial mixes among surveyed countries or other country-specific factors that do not need to be independent of resource-dependence. While debt maturity is slightly longer for the resource firms than other firms in our full sample, the average maturity is shorter in resource-dependent countries than elsewhere. 3.2 Methodological strategy To detect the main determinants for capital structure, we follow Jõeveer (2013), performing an analysis of variance (ANOVA) for three categorical regressors: country, industry and year. We then extend the model using analysis of covariance (ANCOVA) to include continuous firm-specific variables: size, tangibility and profitability. This approach allows us to decompose the variation of dependent variable among the independent variables. The model can be written as YY iiiiiiii =α+ββ jj +γγ kk +δδ tt + θθθθ iiiiii 1 + εε iiiiii, (1) where i,j,k and t are the indexes of firm, country, industry and year, respectively. Yijkt is the capital structure indicator of firm i, country j, industry k and year t. ββ jj is the country fixed effect, γγ kk is the industry fixed effect and δδ tt is the year effect. θθθθ iiiiii 1 presents the firm specific one-period lagged variables and εε iiiiii is the random disturbance. We then extend the model to include the time-varying country-specific factors. The model becomes YY iiiiiiii =α+γγ kk +δδ tt + θθθθ iiiiii 1 + φφφφ jjjj 1 + εε iiiiii, (2) where φφφφ jjjj 1 represents the one-period lagged country-specific variables that can vary over time. We do not include country fixed effects here, as it would capture the resource country indicator. We use pooled OLS to detect the effect of different firm and country specific capital structure determinants. Next, we limit our sample to firms with no close link to the resource sector to determine whether location in a resource-dependent country affects the capital structure of the firm. We use robust standard errors clustered by firm to capture the correlation in regression residuals known to cause bias in OLS estimations using firm panel data (Petersen, 2009). We also cluster standard errors by year to check whether our dummies failed to capture a time effect. 14

15 BOFIT- Institute for Economies in Transition Bank of Finland BOFIT Discussion Papers 10/ 2016 The difference in standard errors is very small compared to pooled OLS with White standard errors, and in line with the capital structure example presented by Petersen (2009). 4 Results 4.1 Variance decomposition In line with Jõeveer (2013), we see the most important determinant of a firm s book leverage is its industry (Table 4). Country is also an important factor. Despite the fact that a major financial crisis hit the global economy during our sample period, year plays a role only in terms of market leverage. Debt maturity structure is clearly more dependent on country of domicile than a firm s industry affiliation. This may reflect the fact that some countries have more market-based financial systems, which coincides with long-term debt, while bank-based financial structures are associated more with short-term debt (Demirgüç-Kunt and Maksimovic, 2002). When we add firm-specific variables, profitability emerges as the most important firmspecific variable in explaining leverage. Profitable firms, not surprisingly, have less need for external debt (Frank and Goyal, 2009). This result is different from Jõeveer (2013), who finds asset tangibility is the most important firm-specific determinant for leverage. Firm size is the most important firm-specific explanatory variable for maturity structure in our data, but our dummies for country and industry remain very important in explaining firm leverage. For columns 7 9 in Table 4, we replace the country dummy with country-specific fixed and time-variant variables. We also add binary indicators for resource firm and resource country. The assigned country variables capture some, but not all, of the variation related to the country dummies in columns 4 6. In particular, the model is poor at capturing book leverage, something expected from the literature (see e.g. Fan et al., 2012). We break this variable down into shortand long-term debt in the regressions to detect variation in detail. Notably, the mere fact of being domiciled in a resource-dependent country appears to be one of the most important country-specific determinants of the level of leverage in our sample firms. The resource firm indicator also explains part of the variation in leverage, even after we control for industry fixed effects. The maturity structure, however, is not explained by our resource indicators when controlling for several other factors. 15

16 Sanna Kurronen Natural resources and capital structure Table 4 Variance decomposition Book Leverage Market Leverage Maturity Book Leverage Market Leverage Maturity Book Leverage Market Leverage Maturity Country 0,37 0,43 0,73 0,25 0,32 0,46 Industry 0,62 0,50 0,27 0,30 0,36 0,18 0,34 0,28 0,19 Year 0,01 0,07 0,00 0,00 0,05 0,00 0,01 0,03 0,00 Size 0,07 0,04 0,26 0,07 0,04 0,22 Tangibility 0,16 0,08 0,10 0,18 0,09 0,11 Profitability 0,22 0,26 0,00 0,29 0,30 0,00 Resource firm 0,01 0,01 0,00 Resource country 0,02 0,03 0,00 Private credit 0,01 0,02 0,03 Market activity 0,01 0,05 0,05 Concentration 0,02 0,01 0,04 Deposit insurace 0,01 0,01 0,01 Corruption 0,01 0,01 0,10 CPI 0,00 0,01 0,02 Profit tax 0,01 0,01 0,01 Common law 0,01 0,05 0,07 GDP growth 0,00 0,02 0,02 Developed 0,00 0,01 0,06 Rating 0,01 0,04 0,06 R2 0,13 0,23 0,35 0,21 0,32 0,39 0,17 0,30 0,37 Obs Notes: Each cell represents the variation that is addressed to the given explanatory variable as a share of total variation explained by the model. Dependent variables: Book leverage Total long- and short-term interest bearing debt to total assets; Market leverage Total long- and short-term interest bearing debt to market value of the firm; Maturity Long-term debt total debt. Independent variables: Size Natural logarithm of assets in US dollars, millions; Tangibility Fixed assets to total assets; Profitability Cash from operations to total assets; "Resource firm" Binary variable for 1=resource firm; "Resource country" Binary variable with 1=Resource-dependent country; Corruption Corruption, high value indicates more corrupt, CPI Consumer price inflation, %, year-on-year; Concentration The share of assets of the three largest banks of total bank assets; Private credit Domestic credit to private sector, % of GDP; Market activity Stock market turnover, % of GDP; "Common law" Binary variable with 1=common law legal origins; GDP growth Annual real GDP growth rate, %; Developed Binary variable with 1=developed country; "Deposit insurance" Binary variable for deposit insurance with 1=deposit insurance scheme; Profit tax Profit tax, % of commercial profits; "Rating" S&P country credit rating in numeric scale. 4.2 Regression results Our regression results presented in Table 5 show that resource firms and firms in resource-dependent countries tend to have less debt, even when controlling for firm- and country-specific factors. The result is especially clear in the case of short-term debt. The coefficient for debt maturity is positive, but insignificant, for both resource indicators. Firm-specific control variables are similar to the main findings of the previous literature. Bigger and more tangible firms have more debt and that debt carries longer maturity. Profitability is negatively associated with leverage. 16

17 BOFIT- Institute for Economies in Transition Bank of Finland BOFIT Discussion Papers 10/ 2016 Table 5 Pooled regression results Dependent variable: Book leverage Market leverage Short-term debt Long-term debt Maturity Size 0.011*** 0.010*** * 0.013*** 0.039*** (,001) (,002) (,001) (,001) (,002) Tangibility 0.135*** 0.150*** 0.125*** 0.277*** (,015) (,016) (,011) (,023) Profitability *** *** *** *** ** (,029) (,038) (,016) (,016) (,031) Market-to-book ** *** *** * *** (,002) (,003) (,001) (,002) (,004) Resource firm *** *** *** *** 0,047 (,020) (,020) (,011) (,016) (,034) Resource country ** *** *** -0,007 0,021 (,010) (,011) (,005) (,007) (,016) Private credit 0.037*** 0.059*** 0.035*** *** (,009) (,010) (,005) (,013) Market activity * *** *** 0.019** (,006) (,006) (,002) (,009) Concentration *** *** *** *** (,015) (,018) (,009) (,022) Deposit insurance ** *** *** -0, * (,011) (,011) (,005) (,007) (,016) Corruption ** *** 0.013*** *** *** (,008) (,009) (,004) (,005) (,013) CPI 0, *** 0.001*** *** *** (,001) (,001) (,001) (,001) (,001) Profit tax 0, * -0, *** 0.085*** (,020) (,023) (,010) (,016) (,032) Common law *** *** *** 0.017*** 0.089*** (,009) (,010) (,004) (,006) (,014) GDP growth 0,001-0, *** 0, *** (,001) (,001) (,0) (,001) (,001) Developed -0, ** -0,004-0, *** (,012) (,013) (,006) (,008) (,018) Rating *** *** *** *** ** (,001) (,002) (,001) (,001) (,002) Constant 0.190*** 0.430*** 0.130*** 0.063** 0.454*** (,036) (,045) (,015) (,027) (,060) Observations 14,457 14,261 16,620 16,457 14,457 R 2 0,19 0,34 0,23 0,29 0,39 Notes: Robust standard errors clustered by firm below coefficient in parenthesis. Year and industry dummies included in all regressions. *p<0.1; **p<0.05; ***p<0.01. Dependent variables: Book leverage Total long- and short-term interest bearing debt to total assets; Market leverage Total long- and short-term interest bearing debt to market value of the firm; St debt Short-term interest bearing debt to total assets; Lt debt Long-term interest bearing debt to total assets; Maturity Longterm debt to total debt. One period lagged values of independent variables are used. Independent variables: Size Natural logarithm of assets in US dollars, millions; Tangibility Fixed assets to total assets; Profitability Cash from operations to total assets; "Resource firm" Binary variable for 1=resource firm; "Resource country" Binary variable with 1=Resourcedependent country; Corruption Corruption, high value indicates more corrupt, CPI Consumer price inflation, %, year-onyear; Concentration The share of assets of the three largest banks of total bank assets; Private credit Domestic credit to private sector, % of GDP; Market activity Stock market turnover, % of GDP; "Common law" Binary variable with 1=common law legal origins; GDP growth Annual real GDP growth rate, %; Developed Binary variable with 1=developed country; "Deposit insurance" Binary variable for deposit insurance with 1=deposit insurance scheme; Profit tax Profit tax, % of commercial profits; "Rating" S&P country credit rating in numeric scale. Independent variables "Tangibility", "Marketto-Book", "Private Credit", "Market activity" and "Concentration" removed from the regressions when the coefficient is not statistically significant at 10% level due to large amount of missing observations. 17

18 Sanna Kurronen Natural resources and capital structure A country s institutional environment matters greatly for firm capital structure. Previous research shows banks tend to provide shorter term debt than debt markets. Our regression here also back up the notion that a higher level of bank credit to private sector is linked to more, but shorter, term debt. Correspondingly, higher stock market activity coincides with less debt and of longer maturity as firms in more market-based financial systems rely more heavily on equity finance and bond issues to raise money. Bank concentration is related to less debt, especially long-term debt. Common law legal origins and deposit insurance schemes are related to less debt and debt with longer maturity. Country credit rating is negatively related to leverage, even if we do not control separately the development level in our regressions. That result is in line with Jõeveer (2013) and could reflect the finding of Fan et al. (2012) that government bond markets seem to crowd out firm debt. In our regressions, the level of economic development is positively related to market leverage. Somewhat surprisingly, debt maturity is shorter in developed countries, which contradicts the positive correlation observed between the two variables in Table 2. Overall leverage is lower in more corrupted countries and debt maturity tends to be shorter. Contrary to our result, Fan et al. (2012) find that the level of debt is higher in more corrupted countries. They reason that this is due to the widespread use of equity financing in less corrupted countries. However, we also have opposite signs for the coefficient when looking at short- and long-term debt in isolation. The association of higher corruption to more short-term and less long-term debt is in line with results of Fungáčová et al. (2015). In countries with weak institutions, banks seem unwilling to provide long-term financing. Similarly, higher inflation coincides with shorter debt maturity. However, as noted from correlation matrix in Table 2, corruption is also highly correlated to development level and country credit rating, so variables are susceptible to multicollinearity that can lead to instability in the coefficients without compromising the model. When it comes to short-term debt and total debt relative to assets, our model seems to capture only about a fifth of variation. In contrast, long-term debt and debt relative to firm value are better captured by our model. This level of explanatory power is in line with previous research with similar cross-country firm leverage data (Fan et al., 2012). Our results are not driven only by flight to quality in the exceptional time of global financial crisis; the results hold for 2007 before the financial crisis hit. Given that we do not have country dummies in our regressions, we confirm that the results are not driven by individual 18

19 BOFIT- Institute for Economies in Transition Bank of Finland BOFIT Discussion Papers 10/ 2016 countries either, by removing one by one countries with a large amount of observations, namely the US, Indonesia, Thailand and China. The results remain robust. 4 Table 6 Pooled regression results with interaction terms Dependent variable: Book leverage Market leverage Short term debt Long term debt Maturity Size 0.015*** 0.014*** -0, *** 0.043*** (,002) (,002) (,001) (,001) (,003) Tangibility 0.155*** 0.169*** 0.151*** 0.281*** (,018) (,019) (,013) (,025) Profitability *** *** *** *** * (,034) (,047) (,019) (,018) (,034) Market-to-book ** *** *** *** *** (,002) (,003) (,001) (,001) (,004) Resource firm 0,056 0,026-0, *** 0.142** (,034) (,035) (,016) (,028) (,061) Resource country ** *** *** -0,007 0,024 (,010) (,011) (,005) (,007) (,016) Private credit 0.040*** 0.063*** 0.036*** *** (,009) (,010) (,005) (,013) Market activity ** *** *** 0.017* (,006) (,006) (,002) (,009) Concentration *** *** *** *** (,015) (,018) (,009) (,022) Deposit insurance ** *** *** -0, * (,011) (,011) (,005) (,007) (,016) Corruption * *** 0.013*** *** *** (,008) (,009) (,004) (,005) (,013) CPI 0, *** 0.001*** * ** (,001) (,001) (,001) (,001) (,001) Profit tax 0, * -0, *** 0.086*** (,020) (,023) (,010) (,015) (,032) Common law ** *** *** 0.018*** 0.091*** (,009) (,010) (,004) (,006) (,014) GDP growth 0.001* -0, *** 0, *** (,001) (,001) (,0) (,001) (,001) Developed -0, ** -0,005-0, *** (,011) (,013) (,006) (,008) (,018) Rating *** *** *** *** ** (,001) (,002) (,001) (,001) (,002) Size*Resource firm *** *** -0, *** *** (,003) (,003) (,001) (,003) (,005) Tangibility*Resource firm -0,035-0, *** 0,058 (,037) (,039) (,025) (,067) Profitability*Resource firm 0, *** 0,039 0,009 0,022 (,055) (,066) (,036) (,030) (,078) Constant 0.150*** 0.392*** 0.128*** 0, *** (,037) (,046) (,015) (,027) (,061) Observations 14,457 14,261 16,620 16,457 14,457 R 2 0,20 0,34 0,23 0,30 0,39 Notes: Robust standard errors clustered by firm below coefficient in parenthesis. Year and industry dummies included in all regressions. *p<0.1; **p<0.05; ***p<0.01. Dependent variables: Book leverage Total long- and short-term interest bearing debt to total assets; Market leverage Total long- and short-term interest bearing debt to market value of the firm; St debt Short-term interest bearing debt to total assets; Lt debt Longterm interest bearing debt to total assets; Maturity Long-term debt/total debt. One period lagged values of independent variables are used. Independent variables: Size Natural logarithm of assets in US dollars, millions; Tangibility Fixed assets to total assets; Profitability Cash from operations to total assets; "Resource firm" Binary variable for 1=resource firm; "Resource country" Binary variable with 1=Resource-dependent country; Corruption Corruption, high value indicates more corrupt, CPI Consumer price inflation, %, year-on-year; Concentration The share of assets of the three largest banks of total bank assets; Private credit Domestic credit to private sector, % of GDP; Market activity Stock market turnover, % of GDP; "Common law" Binary variable with 1=common law legal origins; GDP growth Annual real GDP growth rate, %; Developed Binary variable with 1=developed country; "Deposit insurance" Binary variable for deposit insurance with 1=deposit insurance scheme; Profit tax Profit tax, % of commercial profits; "Rating" S&P country credit rating in numeric scale. Independent variables "Tangibility", "Market-to-Book", "Private Credit", "Market activity" and "Concentration" removed from the regressions when the coefficient is not statistically significant at 10% level due to large amount of missing observations. 4 Regression results for 2007 and the regression results excluding one-by-one United States, Indonesia, Thailand and China are available on request. 19

20 Sanna Kurronen Natural resources and capital structure We test the interaction of resource firm indicator with firm size, tangibility and profitability with the results presented in Table 6. Larger resource firms have less debt and shorter maturity debt than smaller resource firms. More profitable resource firms have a higher level of market leverage. When the coefficient for the size variable and the resource firm-size interaction term are summed up, size does not seem to be associated with higher leverage for resource firms. This finding directly contradicts the very clear result in the earlier literature of a positive correlation between firm size and leverage (Frank and Goyal, 2009). As our results could reflect a strong positive correlation between size and profitability of resource firms, we test for this. While the correlation is higher in case of resource firms than all firms in our data presented in Table 2, the Pearson correlation coefficient of 0.18 it is not high enough to disturb the result by multicollinearity. We also find no evidence that investment intensity of resource firms declines significantly with size. Our results suggest that firms domiciled in resource-dependent countries have less debt, especially short-term debt. This could, of course, be due to the fact that, even when industry fixed effects are controlled for in our regressions, resource firms and firms closely linked to resources in general take on less debt which steers the average financial structure of the resource-dependent country where resource firms play a big role. There are many challenges in finding the right control group when seeking additional evidence that location in a resource-dependent country affects the capital structure of a firm. Many industries such as transportation and certain types of manufacturing are likely to be closely linked to resource firms in resource-dependent countries. Such close relations could affect access to finance for such firms. We limit the sample to two consumer sectors in the data: Consumer Staples and Consumer Discretionary. We expect the consumer sectors to be less linked to resource sector than many other industries. Consumer sectors are not likely to be involved with mineral extraction supply chains, and even if the consumer sectors serve the employees of resource firms, the resource sector is not usually a major employer in a country. 5 Moreover, this control group is sufficiently large (7,541 firm-year observations, of which 1,236 are from resource-dependent countries). The average debt maturity for these firms is 0.48 and book leverage is 0.23, so these firms have less debt and the debt has shorter maturity than that of non-resource firms in general (see 5 Employment data from the International Labour Organization database for Australia, Chile, Colombia, Egypt, Kazakhstan, Norway, Peru, Russia, Saudi Arabia, South Africa, United Arab Emirates and Venezuela show that, on average, mining and quarrying activities account for 1.5% of total employment. 20